Integrating chronological aging and asynchronous aging for enhanced biological age prediction using artificial...

Area of Science:

• Biogerontology
• Artificial Intelligence
• Biostatistics

Background:

• Accurate biological age (BA) estimation is crucial but limited by chronological age (CA) and individual aging variations.
• Existing methods do not fully capture the heterogeneity in aging trajectories.

Purpose of the Study:

• To develop a unified AI framework for BA prediction that integrates CA with a quantifiable measure of asynchronous aging.
• To evaluate strategies for combining CA and asynchronous aging index (AAI) for enhanced BA estimation.

Main Methods:

• Quantified asynchronous aging index (AAI) using a pre-training framework, defined as deviation from a healthy reference population's predicted age.
• Proposed and evaluated three strategies: AAI-score, Loss(AAI,MSE) hybrid loss function, and AAI-driven data cleaning.
• Applied the framework to arterial stiffness data from over 36,000 individuals.

Main Results:

• Integrated approaches uniformly enhanced predictive accuracy compared to traditional frameworks.
• The AAI-score strategy significantly reduced Mean Absolute Error (MAE) in both males and females.
• Area Under the Curve (AUC) for health risk classification substantially improved across both sexes.

Conclusions:

• Asynchronous aging is a fundamental component of biological age.
• Integrating AAI with CA provides a more accurate, interpretable, and clinically promising method for BA estimation.
• The proposed AI framework offers a significant advancement in understanding and predicting individual aging processes.